By Vadim Trifel
I did not comprehend that ChatGPT could fabricate cases,” the attorney rationalized his decision-making, before being reprimanded.1 As every lawyer knows, citing a case without, at a minimum, a cursory review is sacrilege. But that’s old news. We know the lesson: We can’t trust AI to do our legal work for us.
Still, the threat looms. Hollywood writers were on strike until recently, in large part due to the fear that AI will, one day, take over their jobs and write the new hit dystopian series about grownups playing children’s games with life-and-death stakes, or the next generation of legal dramas.2 After all, if there’s one thing that AI excels at, it is fabrication.
Then, a few weeks ago, I received an email about a job offer. It was an invitation to train a legal AI. The email did not go into detail. I assumed I would be involved in some type of machine learning. I would sit behind a computer and correct the program’s mistakes, give it a human touch. And perhaps, unwittingly, I would teach it enough of what I have spent over a decade learning for it to take over my job one day in the not-too-distant future.
In exchange, I would receive $40 an hour. I didn’t respond. No doubt others have and will. And, eventually, lawyers will not be reprimanded, but rewarded, for using their AI friends.
Flash forward 20 or so years.
Big Law Firm (BLF) just upgraded to the latest version of the AI legal assistant, N.O. Mercy 5.0. This AI can review the entire file and compile a summary judgment brief in just a few minutes. But that’s not all. N.O. Mercy is the legal equivalent of Rosey the Robot. It does document review, legal research, and deposition prep—all for the very low price of $30,000 per month. What role is there for associates? Well, there really isn’t one anymore. BLF fired all its associates many years ago. It’s actually cheaper this way.
Judy Bloom, a partner at BLF, looks out of the window of her downtown office at the small dots below. They’re still out there, the protesting law students. They demand employment. She takes a drag from her e-cigarette—which is just as unhealthy as it was two decades ago.
Judy walks over to her desk. N.O. Mercy has finished its latest summary judgment memorandum. She prints out the brief. It is a thick stack of paper. There used to be a day, many years ago, when she would read the entire memorandum. She briefly recalls the time that an attorney was sanctioned for neglecting to doublecheck the work of an AI. It seems like an eternity has passed since then.
In the hundreds of memoranda that she’s read, she hasn’t encountered a single mistake. Not one incorrect citation, nor an inapposite case, nor a misstated fact, nothing. She skims it. It is better than anything a human lawyer could compile, the equivalent of months of work—really, a masterpiece.
But she really should read it over, shouldn’t she? Judy thinks it over. There’s simply no point. And with AI doing so much of the legal drudgery, BLF has done away with hourly fees (you cannot bill for three minutes of work; you’d go out of business). Flat and contingency fees have become a lot more lucrative. So it would actually be more cost-effective not to read it.
Down the street is the Little Law Firm. Sue Little is a talented trial lawyer, but could hardly afford the monthly price tag (even with the small firm discount) for using N.O. Mercy. She doesn’t want to anyway. It’s cheating, she thinks. Plus there’s no joy in it. Sue wants to do it the old-fashioned way: By herself. Create a case from scratch. She relishes the mental challenge.
The case, Ma and Pop v. XYZ Corp., is venued in Hennepin County District Court. N.O Mercy has followed the Rules of General Practice to a T.
Rule 115.05 provides, in part:
No memorandum of law submitted in connection with either a dispositive or nondispositive motion shall exceed 35 pages, exclusive of the recital of facts required by Minn. Gen. R. Prac. 115.03(d)(3), except with permission of the court.
XYZ Corp.’s summary judgment memorandum is exactly 35 pages long, exclusive of the recital of facts. But, in total, the brief is nearly 200 pages.
How could a simple breach of contract case be so fact-intensive? But the fact section is mesmerizing, written in the style of Virginia Woolf, one of the judge’s favorite authors. And the argument cites nearly 70 cases, including some of the judge’s own opinions. It even appears to imitate the judge’s reasoning. Someone who didn’t know any better would be certain that the argument portion of the memorandum was written by the judge, with literary flair.
Sue knows that she has a good case. But XYZ’s fact section is too elaborate—and, she has to admit, too compelling. She has trouble developing a counter-statement of facts. She begins to doubt her own case. And then there’s the argument. It is well-reasoned and includes every conceivable affirmative defense: unclean hands, estoppel, waiver, laches, etc. Sure, plead them (if you can), but you don’t include them all in the argument. But the thing is, somehow, they all appear to have merit.
She spends her nights and weekends researching and writing—80, 90, 100 hours just to respond to this treatise couched as a memorandum of law. Ma and Pop are not happy having spent tens of thousands of dollars. Sue takes a look at Rule 115 again. There’s only so much time: Just two weeks. She doesn’t even have time to answer the phone. It takes all of her energy, but she finally does it. Her brief is finished.
The judge thoroughly reviews both briefs. At the end, he does not know if N.O. Mercy is wrong. No one does. XYZ’s argument is just so convincing that there is no disputing it. Moreover, the judge can no longer handle the caseload, especially in situations when both parties use AI to write their memoranda. And, frankly, he no longer finds passion in his work. What else is there to do after justice has been “perfected”?
Worse yet, there is a new judge running against him in the election. The creators of N.O. Mercy’s slightly modified invention is the new A.I. judge, Judge E. Mental. The voters have spoken: Human error is what creates injustice. They’re convinced that Judge E. Mental will be much more fair and impartial than any human being.
VADIM TRIFEL is a solo practitioner and civil litigator. vadim@trifel.law
Notes
1 https://www.nytimes.com/2023/06/08/nyregion/lawyer-chatgpt-sanctions.html
2 https://www.pbs.org/newshour/show/why-artificial-intelligence-is-a-central-dispute-in-the-hollywood-strikes